[MUSIC] Well, what I want to do today is review a little bit about where we have been, and the kinds of concepts that I've tried to show you in various areas in personalized medicine using the cases. So one of the very first things that people think about when we think about the concept of personalized medicine is using genetic information or other information that we have about a patient to pick the best drug and pick the best dose of that drug for the patient. And so, one of the big areas in which people are thinking about using personalized medicine is in choosing drug therapies appropriately for patients. Why do we have variability in drug action, is one of the questions that we've been addressing and there's variability, as I've talked about before, in drug metabolism, in the targets with which drugs interact to cause their effects. And in other kinds of variability, and the biology in which the drugs act, and one of the questions we have to ask ourselves is, when is this clinical useful? So this is a slide that I showed very early on during this course, and the question is, out of a population of patients when you're dealing with a specific patient, the question is not what happens to the population, but what's going to happen to that particular patient, when you expose them to the drug, would they have a great response? Would they have no response? Would they have some kind of little side effect or will they have some catastrophic side effect? And as I reviewed with you before they're many many reasons that we have variability in drug action. Age, sex, ancestry are part of it. Did we get the diagnosis right? Is the disease that is being treated or some other disease that the patient has, because patients often have more than one disease. Affect how the drug works, is the patient taking many drugs that might interact with each other, are there other things in the environment that might affect the way the drug works? Diet, pollution, stress, those sorts of things. A big deal is whether the patient actually takes the medicine. If the patient doesn't take the medicine, the medicine's not going to do very much. Did the patient actually get the medicine prescribed at the right dose and at the right time? And then there's genetics, so there are many, many factors that go into variability of drug action. Genetics is one of them. I'm showing you this before, this is a slide that shows the distribution of CYP2D6 genotypes across a population. This happens to be a Caucasian population because the distributions are different, depending on ancestry. Again 7% of the populations some are poor metabolizers. They're compound heterozygous or homozygous for loss of function alleles in this gene. And then there's this small group of people who are ultra rapid metabolizers because they have more that one copy, more than one functional copy of the gene. So this raises this idea, this idea of high-risk pharmacokinetics. And one scenario for high-risk pharmacokinetics is, you have a Pro-drug, and it has to be bioactivated, and it's bioactivated through a single pathway. And if that single pathway is deficient because of poor metabolism or because of an inhibiting drug, then the drug won't do what it's supposed to do. Tamoxifen and codeine are examples of pro drugs that require bioactivation through CYP2D6. Clopidogrel which I've talked about also requires bioactivation through CYP2C19. The other high risk for pharmocokinetic scenario is when the parent drug is active but it is metabolized, it's bio-inactivated through a single pathway. And if that single pathway is absent because of poor metabolism and inhibiting drug, or perhaps the pathway is renal excretion and the patient has kidney failure. Then you'll have the scenario with the parent drug will accumulate, and you will get excess drug effect. Warfarin, azathloprine, many 2D6 substrates, particularly those used in neuropsychiatry, fall into this category. Notice here I've used the term narrow therapeutic index and I want to go back and actually talk about what the therapeutic index is for a second. So when we give a drug, when we give a dose of a drug to a group of patients we can actually create what's called a dose response curve. With increasing doses we see more and more patients responding, and this is a efficacy dose response curve. And you can see that as the doses increase the number of patients responding increases. And you can say, well the dose or concentration producing 50% efficacy is marked by that blue arrow there. Now when we give increasing doses of a drug to a large group of patients, we'll also expect to see side effects at some point. Now here's the dose concentration curve those response group for toxicity or for adverse events for this particular drug. Now you can see that there's a very wide separation between the doses required for efficacy in this population, and the doses required for toxicity. We talk about that as a wide therapeutic index drug but a great drug would be a drug that has very few side effects. The side effects that occur are nuisance side effects they're not life threatening and maybe they don't occur in everybody. So a wide therapeutic index drug. That's an example there. What we encounter more often then, and where the genetics part becomes interesting and important, is if those two curves are very close to each other. Notice the way I've drawn it here, not every single patient gets efficacy, but everybody at some dose will always get toxicity. And there's a big overlap between the doses or concentrations required to produce efficacy and those required to produce toxicity. So that's what's called a narrow therapeutic index drug. In the case of variable genetics, all that the genetics will do is move you In some way toward toxicity, in a way from efficacy. So it's in the situation where the toxicity is serious and occurs at doses or concentrations quite close to those producing efficacy that we worry about genetic variation. Clopidogrel and warfarin are great examples because the toxicity is serious bleeding and the lack of efficacy is serious thrombosis. So you really want to balance between those two and if the genetics pushes you one way or the other, that can mean the difference between efficacy and toxicity. So that's the concept of the therapeutic index. So there are other ways besides drug metabolism that can affect the way in which we respond to drugs. This is a genome-wide association study, you've seen many of these before. This is with 22 cases and a very large number of controls looking for Stevens-Johnson syndrome with carbamazepine, in this particular example. And recall that Stevens-Johnson syndrome is this really terrible skin and tongue reaction, and this can be fatal. And as you see on the bottom, HLA variants predict Stevens-Johnson syndrome with carbamazepine, as I showed you on the previous slide. And also with this drug Abacavir that's used in anti-retroviral therapy, with odds ratios that are in the hundreds, so that's a huge effect of a single allele that's not related to drug metabolism, it's rather related to the way in which the body responds to the drug, as I've talked about in a module that dealt with HLA. So one of the things that we worry about when we see variable drug actions is the extent to which genetics contributes. And one of the questions that we have to ask ourselves when we see and when we define genetic variability is when is it clinically useful? And I think the answers to that is, it's clinically useful, if getting the dose wrong, or getting the concentration wrong will produce serious toxicity or serious lack of efficacy. And as in the case of drugs like warfarin or clopidogrel, or toxicity with codeine because of ultra rapid metabolism or toxicity with antidepressants because of lack of an anti-depressant effect and worsening depression. And the other part of knowing the genetics is knowing what to do with the information. If we get a genetic variant and we know that's important for the drug. Does that mean we should switch to another drug? Does that mean we should increase the dose or lower the doses? There's lots of work that has to be done after we define a genetic pathway. That defines what to do with that information in the real world. And we'll come back to that in a later module when we talk about implementation. So another area in which I've spent some time talking about the genetics is using genetics to refine the way in which we define disease. So we can take diabetics and treat all diabetics the same, or we can understand that there are certain subtypes of diabetes. Pretty rare right now, but the hope is that we'll do a better job in the future at certain subtypes that respond to particular drug therapies, because they have different mechanisms that drive them to the common phenotype of diabetes. So that's one way in which genetics can become important, in terms of defining drug mechanisms. The other way is, many, many examples that I've shown you In which we use information around the genetics of disease susceptibility and disease causation. To actually identify new therapies that will be useful [COUGH] in patients with specific genetic abnormalities that drive the development of those diseases. So here are some examples that I've shown you before. Across the top are four mouse hearts from mouse model of Marfan syndrome. The one on the left is a normal mouse heart showing the normal ascending aorta. The second one is a mouse heart that has Marfan syndrome and has been treated with nothing no say they have ascending aortic aneurisms as do the patients. Treating the mice with propranolol does not very much to that ascending aortic aneurism, but understanding the role of TGF beta signaling in Marfan syndrome led investigators to think that maybe we should inhibit TGF signalling, using losartan, the angiotensin receptor blocker. And when you do that in those mice, they do not develop the ascending aortic aneurysm. So that means understanding the mechanism of the disease can lead to new therapies. In this case, with a drug that's long established and well understood. The cystic fibrosis example, there are many many subtypes of cystic fibrosis. Some of them due to abnormal function of that protein. Some of them do the normal function of the protein that just never gets to the cell surface. So once you understand the subtypes then you can start to develop specific drug therapies that target the abnormalities for those sub-types. And there's this drug ivacaftor, which is now available for certain types of cystic fibrosis in which the problem is that the channel doesn't gate properly. And then the last example, those hands are the hands of a patient with severe familial hypercholesterolemia, and I've talked a lot about how understanding the mechanisms in that disease led to the development of statin therapies. And is also leading to the development of other therapies that target the lipid pathways using genetic models to understand what the targets should be. So that's ways in which genetics is informing better therapies and better subdividing of patients with common diseases. The other way in which genetics is having an enormous impact is in cancer. We understand now that cancer is a disease of the genome and it takes multiple genetic abnormalities to drive the development of cancer. Some patients have germ line DNA variants, often in tumor suppressor genes and those can accelerate the development of cancer. Understanding those drivers as I talked about before. Is enabling new drug therapies and new classifications schemes that are based not what the pathologist sees under the microscope. But on what the pathologist is able to do with genetic information, with transcriptional information, with protein expression information to subclassify disease and predict what drugs are going to work and what drugs are not going to work. So we talked a little bit about germline variance and BRCA one. Talked about the case of Angelina Jolie. Germline variants that predisposed to colon cancer and the Lynch syndrome. And then the idea that malignant melanoma can be subdivided not by where it arises in the body and whether there's Sun damage there but, what the genetics are and one of the common variances B R A F, BRAF, V600E. And I've talked about how the development of understanding of the role of that particular variant in driving the melanoma phenotype has lead to the development of drugs that specifically inhibit that abnormal kinase activity. And the interesting part is that there are other tumors, hairy-cell leukemia, certain forms of thyroid cancer, that also respond to V600E inhibition, and that turns out to be a great example of how subclassifying disease based on what you can understand from the genome and not what you send to the microscope might lead to new therapies. And then the last area that I want to just review is the distinction that I have made and that many people make between common and rare variation in the genome and the effect that has on the way in which we develop disease or the way in which we respond to drugs. So, I've talked about many examples of familial diseases, they're shown here, Sickle Cell Disease, Cystic Fibrosis, the Long QT Syndrome, Familial Hypercholesterolemia Marfan Syndrome. And in each of those cases, the way in which people went about solving the problem was to study families. So you study families, you identify the gene and then you can put a dot on a graph like this that explains the relationship between allele frequency in a population and effect size. So this dot for these rare familial diseases is way up in the upper left-hand corner. That means the alleles are very, very rare, but when they occur, they have very large effect sizes. Sometimes not 100% penetrants, but pretty close, and they cause very severe diseases. The other end of the spectrum is the Is the genome wide association paradox. So if you have a patient with familial hypercholesterolemia, or a patient with a really, really long QT interval, you can use the family to get to that. But the other way to get to genes that control cholesterol, or the QT interval, is to study large populations. These are two graphs actually from our own bio bank, which I'll talk about In a couple of modules that describe the distribution of LDL cholesterol in almost 50,000 patients, describe the distribution of the QT interval in more than 30,000 patients and you can see there's a normal distribution. So one of the things we can use, techniques like genome wide association to address, is what is it that makes somebody be in one end of that tail of values compared to the other tail. What puts people in the extremes, or, and notice that most people are in the middle. So we can use GWAS to get at that question, and sometimes that informs what happens in the rare diseases. And I've talked about this technique before and I've talked about how we've used GWAS to study hair color and to study atrial fibrillation susceptibility, macular degeneration, the ability to smell urine after eating. To detect the abnormal smell of asparagus in your urine after eating asparagus. So those are all the kind of traits that people have used to have been studied using the genome-wide association study paradigm. And when we do that, we come up with a dot that's at the opposite end of this plot. The dot is now in the lower right-hand corner. The alleles we discovered through genome wide association are generally common, usually with a minor or low frequency greater than 5% because that's the way the experiment is set up. It's set up to study those alleles and the effect size is very, very modest, so modest in fact that people have criticized the technique of genom-wide association because they say, well, you come up with something that increases the odds of me getting a disease by 50% and that's not anything I'm going to do anything about, so what's the point? And I think I've made the point before that you discover new knowledge through this approach as well. So we now have these two extremes, one based on studying families with rare phenotypes, the other based on studying populations. And they've now replaced the dots with the little symbols, but the real question is what's going on in the middle? And that's where the real excitement In modern genetics and as it applies to personalized medicine is really coming. Because we're beginning to understand, I think, and I've given you examples along the way, how rare and common variation together they conspire to cause a phenotype. So for example, there is this story around the gene called NOS1AP, which turns out to be a major modulator of the QT interval in a GWAS study. And you can say well, a variant in NOS1AP produces a two millisecond increase in the QT interval, and that's not interesting or important to me. But it turns out those common variants also determine the penetrance, or the severity of the long QT syndrome among people who have the rare variance that caused that disease. So it's the combination of the two that predispose to getting the severe disease or getting the milder disease. We're very interested now because the technology is enabling in studying rare variance perhaps with an allele frequency of maybe one in a thousand to one in a hundred and there are many examples now of alleles like that that have large effect sizes. One example that I talked about was SOC 30A8 variants and the idea of protecting against diabetes. So by discovering variants that are relatively rare, they're not as rare as the family things and not as common as GWAS, relatively rare with large effect sizes, we may be able to subdivide patients better and we may be able to devise better therapies based on using those genes that the protein product drug targets. And then there's the idea that common variation, particularly, common variation combined with the environment may conspire to produce large phenotypic effects. And of course, the best example there is the environmental influence of drugs so you take a drug and you may have a variable effect due to a common variant in a drug metabolism gene like CYP2D6 or CYP2C19. That variant does nothing to you in the absence of the drug, so it's the drug plus the common variant that conspire to produce the phenotype. So the future for personalized medicine I think will be in the further definition of these areas using newer sequencing technologies in a large population. [SOUND] >> [APPLAUSE]